Systemic Risk in Real Time



Can central clearing parties (CCPs) weather a sudden storm?

CCPs were recently introduced for standard derivatives to remove counter-party risk from trades. Should the buyer or seller default, then the CCP will take over his trade commitment. CCPs thus effectively insure counter-party risk. Traders are therefore expected to continue to trade at times of elevated counter-party risk. A liquidity dry-up is thus avoided and, with it, the potential for an all-out crisis.

Well, only if the CCP itself survives at such times.

Former Fed chair Ben Bernanke was one of the first to point out that financial stability has become critically reliant on the survival of CCPs. They have become "ground zero" of systemic risk.

CCPs guarantee trades in markets that can rapidly change. It is well known that activity in securities markets is extremely clustered. Quiet periods are followed by sudden bursts of activity, lots of trading, and high volatility. Clear skies are followed by sudden storms.

The potential for such storms has only increased now that markets turned electronic. High-frequency traders, for example, can build large positions within a millisecond. It is therefore not surprising that the CPMI-IOSCO, an authoritative international body of central bankers and regulators, recently formally recommended that CCPs monitor their exposures in real time.

But, how exactly should a CCP do that?

In a recent paper, Wenqian Huang and I propose a risk dashboard for CCPs to monitor their exposure in real time. They can do so because the dashboard is based on analytic results and thus does not require heavy-duty computer simulations. The engine that powers the dashboard was developed in earlier papers (Duffie and Zhu, 2010, and Menkveld, 2015).

The dashboard's "main panel" shows what CCP exposure is towards its clearing members (i.e., traders). A set of subpanels reveals the results of a decomposition of exposure changes. This allows whoever monitors the dashboard to observe what the root causes are for a sudden jumps in exposure. These subpanels essentially show two sets of components:

  • Price-related components: Volatility changes, correlation changes, or price level changes.
  • Trade-related components: Position changes in trader accounts, separated into client and house accounts, and the extent to which positions "crowd" into a single risk factor. (Note that lots of trade does not necessarily increase CCP exposure, simply because traders might be trading out of their positions.)

In the paper, we illustrate the dashboard with real-world CCP data and find:

  • The largest five-minute CCP exposure increases are extremely large. (In other words, the distribution of high-frequency changes in CCP exposure has a fat right tail.)
  • Diagnostic analysis of these extreme increases reveals that they are different in nature than "normal" changes. They are driven more by house-account trading, are accompanied by volatility shocks, and exhibit crowding on a single risk factor. For example, a disappointing earnings announcement by Nokia was followed by traders expanding their Nokia positions, not reducing them.

In sum, insuring trades in high-speed markets requires intraday monitoring with appropriate tools — we propose one such tool. After all, a CCP can only weather a sudden storm if it sees it coming.

P.S.: The paper with Wenqian Huang is here.